[ale] Anyone know if this is true?

Ron Frazier atllinuxenthinfo at c3energy.com
Thu Oct 13 13:07:18 EDT 2011


Ahh, very clever.  Hook me in with my own idea.  8-)  I would consider 
it.  I'd also have to consider if I can allocate the 20+ hours it would 
probably take to get ready.  My findings are currently very very 
preliminary.  Also, I'm one of the more newbie Linux users in the 
group.  I'm not sure how qualified I am.  What did you have in mind?

Ron

On 10/13/2011 11:51 AM, Jim Kinney wrote:
>
> So can you consider a presentation of your findinds and perhaps some 
> testing?
> :-)
>
> On Oct 13, 2011 10:52 AM, "Ron Frazier" <atllinuxenthinfo at c3energy.com 
> <mailto:atllinuxenthinfo at c3energy.com>> wrote:
>
>     Hi guys,
>
>     This thread has prompted me to do a bit of research to try to see
>     if I can find a consensus on the swap size issue.  At the moment,
>     it seems like ask 100 people and get 100 different opinions.  I
>     haven't uncovered enough data to tabulate and summarize it at this
>     point.  I'm pretty sure my linux machines have plenty of swap for
>     applications.  I have 8 GB of swap on two machines which have 8 GB
>     of RAM.  If I look at the system monitor program in Gnome, it
>     looks like the swap is rarely ever touched.  By the way, having 8
>     GB of RAM in a laptop is a nice, new, liberating experience for
>     me.  It's really nice to be able to open several dozen browser
>     tabs and a dozen or more applications without the machine even
>     breathing too hard.  This is my first laptop capable of that.  I
>     also have 8 GB of swap on a machine with 4 GB of RAM, so it should
>     be sitting pretty, so to speak.  I haven't found anything that
>     says extra swap is harmful.  What I don't know, is whether the two
>     8 GB machines would be able to hibernate (suspend to disk)
>     properly, if the swap is equal to the RAM.  I may have to increase
>     the swap on those to 10 GB - 12 GB.  This is not an issue in
>     Windows since it uses a separate hibernate file.
>
>     In my research, I found this article (
>     http://lukasz.szmit.eu/2009/11/compcache-swap-on-linux-desktop.html ). 
>     The article is a bit old, but this talks about a fascinating
>     project called Compcache.  Here's a quote from the page:
>
>     ---> quote on <---
>
>     Compcache is an open source project implementing an innovative
>     approach to swap. This has been done before, but not for swap.
>     Users of DOS and early Windows versions will remember
>     DoubleSpace/DriveSpace, which was used to virtually expand
>     available disk space, by storing files in a compressed form.
>     Compcache does exactly that, but for swap, by creating a new block
>     device in the system which interfaces with the special compressed
>     memory region in RAM. On the plus side, Compcache can also be
>     configured to use an alternative swap device when the RAM swap
>     area is full.
>
>     ---> quote off <---
>
>     I think that is a really cool idea for low resource machines. 
>     While I don't know if I'll ever use it, since my modern machines
>     have a decent amount of RAM, it could really benefit older,
>     smaller machines.  For example, I have an old IBM Thinkpad with
>     160 MB (yes MB) of RAM.  I've pretty much retired the machine.  It
>     does run a GUI based version of Linux, just barely, but is
>     painfully slow.  The old 300 MHz processor doesn't help much
>     either.  I think I have an old version of Lubuntu on it.  Anyway,
>     this type of technology could give the machine more breathing room
>     by compressing the memory, so it would be like having 256 MB of
>     RAM.  I also have an old Toshiba laptop which is topped out at 1
>     GB or RAM.  Both Ubuntu 10.04 and Windows XP run pretty well. 
>     However, I could compress 512 MB of RAM and then effectively have
>     512 MB or normal RAM and 1 GB of swap.
>
>     Here's the link for the Compcache project:
>     http://code.google.com/p/compcache/
>
>     Here's an interesting quote from their site: "With compcache at
>     hypervisor level, we can compress /any/ part of /guest/ memory
>     transparently."
>
>     Now, while I'll admit I don't understand all the implications of
>     that statement, it looks like you could essentially compress all
>     the RAM if running a lightweight hypervisor and running your OS as
>     a guest.
>
>     The project website also points out that embedded systems could
>     benefit from the technology, where you have to justify every chip
>     you put into a device.
>
>     Finally, here is an interesting quote from the original article:
>
>     ---> quote on <---
>
>     On my desktop, a Dell Precision S390 with 2GB DDR2 RAM and a
>     Maxtor Diamond Max 9 80GB drive, I am getting the following hdparm
>     results (average of three runs) for my disk swap, and my compcache
>     swap:
>
>         * Swap on disk: 58MB/s
>         * Compcache swap: 557MB/s
>
>     An order of magnitude better bandwidth at no expense? I like that.
>
>     ---> quote off <---
>
>     I like that idea too.  I'd like to know what you guys think of
>     this concept.
>
>     Sincerely,
>
>     Ron
>
>     On 10/13/2011 8:58 AM, Rich Faulkner wrote:
>>     For me depends upon the system as to swap size, but if I plan on
>>     using hibernation features I have swap just over the size of RAM
>>     as in 1-1/2 times as the general (old) rule that I've
>>     followed...generally a couple of gigs for a desktop and I leave
>>     it at that.  Being as I'm only building desktops and laptops
>>     lately I'm not speaking to servers.  An interesting experiment is
>>     to do test installs to various system configs and see what a
>>     given distro will do for a default installation.  I consider this
>>     a benchmark from the developers on an "ideal" configuration given
>>     the hardware provided.    RinL
>>
>>
>>     On Thu, 2011-10-13 at 08:45 -0400, Scott Castaline wrote:
>>>     On 10/12/2011 09:40 PM, Tavarvess Ware wrote:
>>>     >
>>>     >  Scott I read the ram x 2= swap in my Linux classes as well and have
>>>     >  generally followed that, but with memory soaring as it has lately i am
>>>     >  starting to rethink that.  A system 48gigs of memory would be 96 in
>>>     >  swap..... I wonder if te old format has changed and I haven't heard yet.
>>>     >
>>>     I only go RAM x 2 = swap for the 1st 2 GB of RAM so 2GB RAM = 4GB swap.
>>>       From there on it's RAM x 1 = swap so 4GB RAM = 6GB swap. So your 48 GB
>>>     of RAM = 50GB swap, and yup that's one hell of a lot of swap space.
>>>     >
>>>     >  On Oct 12, 2011 9:32 PM, "Scott Castaline"<skotchman at gmail.com  <mailto:skotchman at gmail.com>
>>>     >  <mailto:skotchman at gmail.com>>  wrote:
>>>     >
>>>     >      On 10/12/2011 04:14 PM, planas wrote:
>>>     >      >  On Wed, 2011-10-12 at 15:13 -0400, Geoffrey Myers wrote:
>>>     >      >>  'Just so you all know, when determining how much space to assign to
>>>     >      >>  swap: Swap isn't just used for paging or virtual memory
>>>     >      management; swap
>>>     >      >>  is also used by power management for suspend-to-disk
>>>     >      (hibernation). '
>>>     >      >>
>>>     >      >>  I seriously don't know, so I'm asking.
>>>     >      >>
>>>     >      >
>>>     >      >  I have seen that a good swap size is ~1.5x the RAM.
>>>     >      >  --
>>>     >      >  Jay Lozier
>>>     >      >  jslozier at gmail.com  <mailto:jslozier at gmail.com>  <mailto:jslozier at gmail.com>
>>>     >      >
>>>     >      >
>>>     >      >
>>>     >      I remember from somewhere that upto 2GB use 2.0x RAM above 2GB of
>>>     >      RAM go
>>>     >      with 1:1 ratio so 4Gb RAM = 6GB swap. I don't remember why 2x on the
>>>     >      first 2GB and this goes back to when 4GB was a lot on pre-configured
>>>     >      retail boxes. So like Geoffrey I can't see having 18GB of swap for a
>>>     >      16GB machine.
>>>     >    
>

-- 

(PS - If you email me and don't get a quick response, you might want to
call on the phone.  I get about 300 emails per day from alternate energy
mailing lists and such.  I don't always see new messages very quickly.)

Ron Frazier

770-205-9422 (O)   Leave a message.
linuxdude AT c3energy.com

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.ale.org/pipermail/ale/attachments/20111013/83339503/attachment.html 


More information about the Ale mailing list